The latent failure rate (LFR) test is a reliability qualification test widely used during semiconductor device manufacturing. It is a measure of life cycle of the semiconductor devices. In conventional manufacturing processes, the LFR is typically measured by operating the devices at elevated temperature and voltage for 1000 hours and measuring the failure rate. Using the Arrhenius law, the expected failure rate at operating temperature is calculated.
In the conventional solution the LFR burn-in temperature is estimated and is not actually measured in line. The conventional method of setting LFR burn-in temperature is described. In a first step, the average or maximum power of the device is estimated. In a second step, the junction temperature is computed using the temperature coefficient of the package (θja) which is typically provided by the packaging manufacturers and the ambient temperature as described below:Tj=Power*θja+TambientThe LFR burn-in temperature estimated using the formula above is approximate and can be inaccurate by several tens of degrees. Because of this approximation error in setting the burn-in temperature, in some cases the actual junction temperature can be higher than process allows which can lead to “inter-metallic” growth on the pad, also known as “Purple Plague Failure”. Purple plague failure is an undesirable failure mechanism well known in semiconductor manufacturing. Alternately because of an approximation error, the actual junction temperature may be lower in which case there is danger of under-stressing the parts, leading to weak or unreliable devices being shipped to customers, which is highly undesirable.
It would be desirable to have an accurate and in-line measurement of the LFR burn-in temperature.